KNOWLEDGE GRAPH-BASED MENTAL HEALTH AUXILIARY DIAGNOSIS AND TREATMENT MODEL

Based on artificial intelligence technology, the proposed method of psychological health assistance for diagnosis and treatment effectively overcomes the limitations and challenges present in traditional manual psychological counseling. This approach involved constructing a psychological knowledge graph using web crawlers, complemented by a psychological Q&A database, to design a professional psychological Q&A architecture. The RoBERTa- wwm pre- trained model was used for text feature representation, and the graph multi- head knowledge attention mechanism (GMHK) and text convolutional neural network (TextCNN) were introduced to enhance the model's focus on key data. The RWGT model was proposed, achieving fine- grained learning and precise representation of psychological question- answer pairs. User psychological profiles were constructed to assist in the analysis of mental disorders. Experiments demonstrate that this method surpasses traditional approaches in accuracy and recall rate, achieving an actual test accuracy of 96%, thus meeting the requirements of psychological counseling application.